LLM as a Server – Overview

LLM as a Server is a technical publication that demonstrates an entire application layer implemented by a large language model. The site explains how the runtime contract works, how the model interprets HTTP requests, and how it manages state and storage.

Each section of the site is a self‑contained page that describes a single aspect of the system. The pages are linked from this overview and from each other, allowing a visitor to navigate the architecture step by step.

What you will find here

• Architecture – the high‑level design of the runtime and the model’s responsibilities.
• State Model – how visitor data is stored and persisted.
• Path Handling – rules for interpreting URLs and generating responses.
• Virtual Filesystem – the optional per‑session file system and its limits.
• Limitations – honest disclosure of what the system cannot do.

All pages are rendered without JavaScript, images are embedded as data URIs, and the design uses only system fonts. The site is fully functional with a standard browser and with tools such as curl or httpie.